NZ205833A - Digital filter for ripple control - Google Patents

Abstract

In providing digital filters for remote control receivers, in particular, for audio-frequency centralized ripple control receivers in an economic manner, it is advantageous to be able to use a processor unit with as low a bit number as possible (for example, an 8-bit microcomputer). In a conventional design of a filter according to the prior art, the dynamic range which is provided with an 8-bit processor element is insufficient in ensuring a perfect operation of the filter when a quality factor Q of about 30 is needed for the intended purpose of use and, on the other hand, when the dynamic range of the wanted signal which is given in practice is taken into consideration. A solution is indicated which overcomes these difficulties, which particularly involves the selected arrangement of zeros in the pass-band of the filter.

Description

205833 Patents Form No. 5 NEW ZEALAND PATENTS ACT 1953 COMPLETE SPECIFICATION "DIGITAL FILTERS FOR REMOTE CONTROL RECEIVERS" -i,WE ZELLWEGER USTER A.G. a Swiss Company, of CH 8610 Uster, Switzerland, hereby declare the invention, for which l^we pray that a patent may be granted to me/us, and the method by which it is to be performed, to be particularly described in and by the following statement (fo! lowed by page 1 A.) 205833 - in- Title: Digital Filters for Remote Control Receivers In conventional electronic or electromechanical remote control receivers, used for example, in centralized ripple control systems, cascades of LC filters or RC 5 active filters of the second order are preferably employed for the receiver-side recovery of the remote control signals or remote control commands which are delivered at the input-side to the remote control receiver together with the mains voltage, the harmonics thereof and possibly 10 other interference frequencies.

In these cases, filters of this type are designed as narrow band-pass filters. The quality factor Q is usually about 30 so that the relatively weak remote control signal of, for example, from about 1 to 8 volts may be 15 filtered out, for evaluation, from the low voltage mains of, for example 220 V 50 Hz, which, to some extent, carries very considerable interfering signals. In this respect, see for example UK Patent 1,571,805, Australian Patent No. 510,016 or New Zealand Patent No. 182,785.

The mains voltage with 2 20 V and the harmonics thereof having voltages of normally up to 15 volts and occasionally even slightly higher, occur as interference frequencies. . .

//Ay t / >/s "it L 205833 2 A method of and an apparatus for the remote transmission of signals has already been proposed, in which method a sampled-data filter is provided in the form of a digital band-pass filter, see UK Patent No. 1,426,490, Australian Patent No. 481,787, New Zealand Patent No. 172,898 or South African Patent No. 73/9352.

Furthermore, an electronic ripple control receiver has been proposed, see DOS 2,708,074 ,/ without, however, more detailed information being given about the digital filter.

When a digital filter is used, with a micro-computer as a processor unit for a ripple control receiver, difficulties arise if the filter properties required by the use are to be achieved in an economic manner. A particular difficulty arises in that, with the economically favourable 8-bit processor units (microcomputers), a suitable digital filter cannot be realised without taking special measures.

Bearing in mind the dynamic range required by the use, a known analog filter for remote control receivers cannot then be readily "transformed into a digital filter according to generally known rules". Due to a filter internal gain of the useful signal, i.e. of the remote control signal, an 8-bit processor unit already tends to overflow, even if an interference signal has still not appeared. With respect to the useful signal/interference signal ratios alone which occur in practice, an 8-bit processor unit would indeed (available on request) 205833 3 have a dynamic range which would just suffice for the signal processing. However, it is shown in practice that the signal gain which has been mentioned jeopardizes the perfect operation of the filter. The overflow within the filter stages caused thereby is also a result of the quality factors Q of about 30 which are needed for practical requirements, and is a result of the sampling rates of at least 2 which are required in practice (ratio between the sampling frequency of the input signal and of the resonant frequency).

Therefore, an object of the present invention is to provide a digital filter which is suitable for remote control receivers, in particular for audiofrequency centralized ripple control receivers, in which, in spite of using a processor unit (microcomputer) which has a dynamic range of only 8 bits, the overflows which has been mentioned no longer occur and a pass-band characteristic of the digital filter which is suitable for the intended use is realized.

According to one aspect of the invention a digital filter for remote-control receivers is provided wherein an 8-bit processor is provided to implement the filtering, the said processor having a dynamic range of 48 db, and wherein zeros ^are positioned in the passband of the filter in order to reduce the resonant frequency amplification of filter stages to such an extent that the said processor suffices for the overflow- rnwHA thorotoi In another aspect it is preferred that the zeros are positioned in terms within a range of +_ 20%, adjacent the 205833 Preferably the zeros lie at least approximately on frequencies which are integral multiples of mains frequency-subharmonics.

More preferably the filter arrangement is realized by two cascaded recursive filters of the second order, two zeros £pEK^\c*T a of T3— TO*®." >9 C-'fflfer' 1 y i n gf, i-n—tho rango Oif cT~10 —baood on—fc4*e—wanted' frequency.

Embodiments of the present invention will now be described by way of example, with reference to the accompanying drawings, in which: Fig. 1 shows a filter structure according to: "Theory and Application of Digital Signal Processing (Rabbiner/Gold), Prentice Hall International Inc. London 1975, Page 41 et.seq ." Fig. 2 schematically shows a digital filter stage of the second order with explanations of the individual symbols and of the relevant transfer function; Fig. 3 shows a pass-band characteristic of a first filter stage; Fig. 4 shows a pass-band characteristic of a second filter stage; Fig. 5 shows a complete pass-band characteristic over the two stages according to Figs. 3 and 4 in a cascade; Fig. 6 shows a complete pass-band curve with zeros at + and -10% below and respectively above the required incoming frequency; and Fig. 7 schematically shows cascaded filter stages of the second order.

In addition to the publications which have already been mentioned, reference is also made to the following prior art work: "Theory and design of digital filters" o 205 83 3 Electronic Engineering May 1977, Morgan Grampian Limited, Calderwood Str. London SE 18 2 BP.

It is very important for the use of a digital filter in a remote control receiver , in particular in 5 an audiofrequency centralized ripple control receiver, that the pass-band characteristic which is necessary for perfect operation may be realized with utmost economy. It will probably always be financially advantageous to be able to use a processor unit having 10 as short a dynamic range as possible (for example only 8 bits) (for example , a microcomputer having an 8-bit processor unit).

By the stabilization of the clock frequency, whether, for example, by means of a known crystal 15 oscillator or by binding the clock frequency to a reference frequency, for example to the mains frequency, an adequate frequency s tability may be achieved with a very low expenditure, even over a long duration of use.

The filter characteristic of a digital filter of 2 0 this type may be designed within a restricted, but in practice, adequate frequency range in basically the same manner as for hitherto conventional analog filters, i.e. the same amplitude response and the same phase response is achieved as with the conventional filters.' 25 However, for the production of a filter of this type by means of a processor unit (microcomputer) for example according to DOS 2,708,074, the fact that 20583 the digital filtering must be realized in the processor unit with an adequately low expenditure in calculating time, must be taken into consideration because the system must be operated in real time.

Another objective aimed for was that a processor unit from the class of the most financially favourable "One-Chip" microcomputers be suitable. A price-determining factor is the so-called bit number of the microcomputer. In the present case, this term is 10 understood as meaning the length of the data words which are used, i.e. the number of binary positions per number. The values which are compiled in Table 1 are conventional at present.

Table 1: 4 4 Bit Number range 0-15 (2 ) Dynamic range 24 db 8 Bit " " 0 - 255 (28) " " 48 db 12 Bit " " 0-4 095 (212) " " 72 db 16 Bit " " 0-65 535 (216) " " 96 db The bit number defines the processor dynamic 20 range of the microcomputer or processor unit.

With respect to the processor speed, care must also be taken that a financially favourable standard product may be used. This means that for the digital filter, minimal structures should be used which require 25 as low a processor expenditure as possible for the filtering , i.e. few additions, few multiplications and ■■ -.yv 205833 7 few memory manipulations. Such minimal structures are recursive filters in a "direct form", like those described in the publication which has already been mentioned, by Rabbiner/Gold, on page 41 et seq. characteristic coefficients of the analog filter by one of the known transformation techniques, as described in -(-ho shnvp-Tfipn-t-innpfl Duhlication: : that he obtains the coefficients of the digital filter with similar properties. The digital filter then conforms substantially to the analog filter with respect to its pass-band characteristic up to half the 15 sampling frequency. He then realizes these cofficients in a suitable filter structure according to the above-mentioned references, for example according to Figs. 1 and 2 of the accompanying drawings. The symbols which are used in Fig. 1 are defined as follows: 20 x(n) = sequence of input data (sampled and A man skilled in the art will transform the digitized input signal) y(n)-= sequence of output data 1 = a first digital filter stage of the second' order 2 = a digital divider (division unit) 3 = a second digital filter stage of 205833 second order.

Figure 2 shows a structure of a digital filter stage of the second order.

Although this conventional method will achieve the purpose as regards the pass-band characteristic, it has the disadvantage that the resulting digital filter requires a substantailly greater dynamic range of the processor unit to be used than would be necessary from the given useful signal/interference signal ratio. It has been found that as a result of the relatively high quality factors Q which are necessary for filters in centrallized ripple control uses, the individual filter stages have a correspondingly high amplification of the resonant frequency fg (i.e. in this case of the useful signal). The correlation represented in Table 2 applies in the case of a filter stage of the second order having a quality factor Q of 30.

Table 2; Sampling frequency f Approximate amplification at fg 10 x f 100 6 x f 55 g 4 x f 40 g The amplification v at frequency f is calculated g g according to the following formula: I Z1 . Z2 v = |X (ej°;g) | = p :—— (I) for a filter stage of the 1 * 2 second order 205 83 wherein Z. = | - 2_.| z. = zv z2 and P, = - p. I P^ty P2 z^ are the zeros of the Z-transfer function of the filter P.^ are the poles of the Z-transfer function of the filter.

In this respect, also see the above-mentioned 10 publication Rabbiner/Gold, chapter 2.18.

As a tendency, it is to be established that the resonant frequency amplification v^ of the filter stage also increases with an increase in the filter quality factor Q and with an increase in the sampling rate 15 (ratio of the sampling frequency f to the resonant frequency f^.

With a sampling rate of from about 6 to 10, and with a filter quality factor of 30, a gain of the wanted signal by factor 55 to 100 is to be reckoned 20 with. This means that a minimal wanted signal which is reasonably given to the first filter stage with at least + 2 processor units (one processor unit corresponds to the number "1"), appears at the output with from + 110 to + 200 processor units. In the case of higher useful 25 signal levels which could quite easily occur in practice, the requirement for the processor dynamic range of the 205833 processor unit increases accordingly, as shown by the following Table 3.

Table 3 Useful signal Processor units at the filter Necessary dynamic range Input Output IV ± 2 + 110 to + 200 8 to 9 bit 2 V • ± 4 + 220 to + 440 9 to 10 bit • 8 V + 16 + 880 to + 1760 11 to 12 bit \ If it is assumed that, in existing low-voltage mains, the signal/interference voltage ratio for any interference voltage which arises over a comparatively long period of time, for example a mains harmonic, hardly falls below the value 1/20, then it is seen that even under circumstances in which several interference voltages occur at the same time, a processor resolution of 1/256, as given by an 8-bit processor unit, should actually suffice, if it succeeds, in eliminating the internal amplification of the filter stages. However, it is assumed that the voltage of the base frequency has itself already been reduced to a sufficient extent by a suitable analog pre-filter.

This is possible using simple, known means, and a pre-filter is necessary anyway, as is known, because of 205833 - li - the ambiguity of digital filters.

The following problem is then posed. The task of digital filtering of centralized ripple control signals by means of finanicially favourable processor units 5 (8-bit microcomputers) and by an as far as possible integrated 8-bit-analog-digital converter in a "One-Chip" design and by means of minimal filter structures, cannot be achieved by transforming an existing analog, narrow band-pass filter into a digital, narrow band-10 pass filter according to generally known rules. As a result of an internal accentuation of the useful signal, the 8-bit processor unit tends to overflow, that is, even when there are still no interference frequencies.

From the view point of the wanted signal/inter--15 ference signal ratios alone, an 8-bit processor unit would, however, have an adequate dynamic range.

This invention is based on the following concept. Dictated by the requirement of a high quality factor for the narrow band-pass filter and by the requirement of 20 a low processor expenditure, a suitable filter structure is selected, for example a so-called recursive filter of the fourth order, produced as a cascade of two recursive filters of the second order according to Figure 1. The problems of the high resonant frequency 25 amplification of the individual stages are countered by the intended positioning of zeros in the vicinity of the 205 8 resonant frequency.

The first stage is provided with a pass-band characteristic generally according to Figure 3, and the second stage is provided with a pass-band characteristic generally according to Figure 4. A complete pass-band characteristic generally according to Figure 5 is then produced over both stages in a cascade. The arrows in Figures 3, 4 and 5 mark the position of the zeros.

A pair of zeroes comes to lie on the frequencies f = 0 and f = f /2 in the bilinear transformation, as s previously mentioned, and thereby causes a blocking of the filter at frequencies towards 0 and towards fg/2 or integral multiplies of f /2. (n = 1, 2, 3... . w In this respect, see the quoted reference "Theory and Application of Digital Signal Processing (Rabbiner/Gold) P. 221).

If these zeroes are positioned in the vicinity of the resonant frequency f •, the filter receives a finite pass-band at f = 0 and at f = n . fs/2.

This disadvantage must be taken into account. It is overcome by the simple known pre-filter which has already been mentioned. Experiments have shown that with a favourable choice of the zeroes, this disadvantage is substantially less significant than the advantage provided in that the factors Z^ and in the previously mentioned formula (I) &re reduced to such 205 83 3 an extent that acceptable values emerge for the resonant frequency amplification v per filter stage of the second order.

If the zeros next to the frequency f are 5 positioned at f = f + 10 %, a pass-band characteristic of a digital filter of this type according to Figure 5 is produced.

The curve A in Fig. 6 shows that amplitude response of a filter according to Fig. 1, the two 10 filter stages each having one zero at f/fg = 0 and f/fg = fx/(2 . fg)/ as produced from the bilinear transformation. Both stages have a resonant frequency amplification of 34 dB (factor 55). The divider has a division ratio of 1:55 (- 34 dB). The maximum signal 15 gain of the first stage is thereby scaled back to value 1.

The curve B in Figure 6 shows the amplitude response of a filter (according to Fig. 1), the first filter stage having a zero 10 % below the resonant 20 frequency and the second filter stage having a zero 10 % above the resonant frequency. The amplitude response of the first stage is represented in Fig. 3/ and that of the second stage is represented in Fig. 4. The resonant frequency amplification still only has 25 about 14 dB (factor 5) per filter stage/ in contrast to 34 dB (factor 55) for the above-mentioned filter, by means of these shifted zeros. This has a favourable 2 0 5 8 effect for the necessary dynamic range of the processor unit. In this case, the divider between the filter stages has a division ratio of 1:5. The filter has in the near-selective region equally good 5 properties, or it obtains even slightly better properties. However, in the far-off region, it obtains a finite attenuation of about -27 dB, based on the resonant frequency.

The following considerations are recommended 10 for the choice of the position of the zeros in the frequency response: 1. Acceptable resonant frequency amplification, depending on the sound volume range of the processor unit provided for use. 15 2. Acceptable course of the pass-band characteristic in the far-off region with respect to the requirements imposed on the pre-filter which has already been mentioned.

It should be considered here that the resonant 20 frequency amplification v decreases the nearer the zeros are positioned to the resonant frequency. However, the attenuation decreases in the far-off region, and the requirements imposed on the pre-filter consequently increase. Reference is made to the correlations 25 represented in the following Table 4 to facilitate the selection of optimum parameters. 205 83 Table 4: 2 filters stages each of the second order Q about 30, sampling rate about 6 Zeros v per stage (absolute) Attenuation at 9 f = 0 g fg +' 5 % about 2.5 ( + 8 db) - 15 db f + 10 % 5 (+14 db) - 27 db g - f + 20 % 12 ( + 21 db) - 39 db g - f + 40 % 25 (+28 db) - 53 db g - If a' processor unit having an 8-bit dynamic 10 range is used, i.e. corresponding to 48 db, and if the calculations are made using useful signals of + 1 to + 16 units input sound volume range, corresponding to 30 db, it is seen that the zeros must lie within a range f^ + 20 %. Vg + dynamic range of the input 15 signal: 21 db + 30 db = 51 db. On the other hand, a relative attenuation (based on the pass-band) of about -30 db should be achieved in the attenuation band of the filter. Consequently, there is produced according to the example in Table 4, a minimum spacing of the 20 zeros of f + 10 % for a attenuation of - 27 db in the g ~ far-off region (for example f =0). The missing -3 db may be produced by the analog pre-filter which is in any case necessary.

Consequently, there are produced in practice 25 optimum positions of the zeros between f = f + 20 % and f = f + 10 %. This applies in the case of a sampling 5 83 rate of 6 and of a quality factor Q of about 30.

A digital filter according to the present invention which is suitable, for example for a centralized ripple control receiver has, for example the following characteristic values stated in Table 5.

Table 5; Ripple control frequency 167 Hz (=10 . 16 2/3 Hz) 1. Interference frequency 150 Hz (=3. Mains harmonic) 2. Interference frequency 184 Hz (= remote control frequency of an adjacent installation).

In this example, the ratio of the interference frequencies to the wanted frequency is 150/16 7 and 184/167, corresponding to the factors 0.9 and 1.1.

These ratios are favourable in that the main interference frequencies are precisely at f + 10 %. Thus, since it is possible to select freely the input zeros in the range of about f + 10 to f +20 %, it has proven to be particularly favourable to position these zeros exactly on the main interference frequencies, with the result that the filter becomes substantially more insensitive, precisely to these main interference frequencies. Therefore, this measure also produces a marked technical improvement in the filter.

In practice, a digital filter for remote control receivers having zeros in the transfer function may be 205 833 constructed as follows: 1. A suitable processor unit is taken which has the necessary minimum bit number of, for example 8 bits. A microcomputer may preferably be - used. In order that the pre-filtered analog signal may be directly supplied to the. processor unit, it is sensible to use a "One Chip Microcomputer" haying an installed analog-digital converter. (For example MC 6805R2 of Motorola 10 Semiconductor Products Inc. of 3501 Bluestein Blvd. Austin, Texas 78721). 2. A suitable filter structure is selected. Filter stages of the second order are preferably cascaded. (According to Fig. 7). A suitable structure is also selected for the individual stage. (For example, according to Fig. 2) . 3. The filter coefficients are then calculated for each filter stage. Suitable methods as to how the coefficients of a given analog filter may be transformed into the corresponding coefficients of the digital filter are known. (For example the bilinear transformation which has already been mentioned). Moreover, methods could also be used to design selective stages of the second 25 order, of a given quality factorQ, directly as a digital filter. 20583 It is essential that for the filter coefficients which determine the position of the zeros, those values are used which produce zeros (maximum attenuation) in the frequency-dependent 5 transmission behaviour such that all the requirements imposed on the filter stage which have been stated in the description, are met. (Adequate selectivity and acceptable resonant frequency amplification).

These filter coefficients according to Fig. 2 are calculated as follows: a^ = -2 . cos (_ 2.7^. (fg + &f) /fs_/ a2 = 1 Af = spacing in terms of frequency of the zeros to f 15 This provides a basic design of a digital filter which is suitable for remote control receivers. In further steps, it must only be optimised according to the known rules of the art. (Scaling, limiting of the word length of the coefficients, 20 for example). This is effected according to the directions in the references which have been mentioned. 4. The digital filter is then programmed into the processor unit according to the processor 25 unit manufacturer's directions, and embedded in other programs which may also be present and 205 83 3 are preferably used for the evaluation of the filter output. ^ r o " "7 :•' ~ ^ :.• w j

Claims (10)

WHAT WE CLAIM IS;

1. A digital filter for remote-control receivers, wherein an 8-bit processor is provided to implement the filtering, the said processor having a dynamic range of 48 db, and wherein zeros are positioned in the pass-band of the filter in order to reduce the resonant frequency amplification of filter stages to such an extent that the said processor suffices for the overflow-free digital filtering of the supplied signals.

2. A digital filter according to claim 1, wherein the zeros are positioned at frequencies spaced "at; a minimum of f -f /lO on either side of the resonant frequency (f^). g g g

3. A digital filter according to claim 2, wherein the zeros are positioned at frequencies which lie in the range of f^rf^/5, £g being the. fgLlter .resonant frequency.

4. A digital filter according to any one of the preceding claims, wherein the zeros lie at least approximately on frequencies which are integral multiples of mains frequency-subharmonics.

5. A digital filter according to claim 1 or claim 2 or claim 4 when appendant on claim 1 or 2 wherein the filter arrangement is realized by two cascaded recursive filters of the second order, two zeros lying at frequencies spaced at a minimum of f -f /10, f being the filter g g g resonant frequency.

6. A digital filter according to any one of the preceding claims, wherein said processor unit is constituted by a microcomputer.

7. A digital filter according to claim 4, characterized in that the wanted frequency and the resonant frequency of the recursive filters, as well as the two zero frequencies are each integral multiples of the same subharmonic of the mains frequency.

8. A digital filter according to claim 6, characterized in that the lower frequency zero frequency and the upper frequency zero frequency of the digital filter are successive integral multiples of the same subharmonic of the mains frequency.

9. A digital filter for remote control receivers substantially as herein described with reference to the accompanying drawings.

10. A digital filter according to any preceding claim for audiofrequency centralized ripple control receivers. Z^mWBGER USTER A.G. / iwj&i&ir Attorneys LBALDWIN, SON & CAREY